Introduction

When animals compete for resources, their competitive abilities and behavioural strategies can be expressed as changes in movement activity. This activity has been positively correlated with higher growth rate, larger size and shorter development (Werner and Anholt 1993), as more active animals frequently encounter more food (Olsson et al. 2002; Brodin 2008; Liebgold and Dibble 2011). The levels of activity can be explained as predictable animal behaviour in response to changes in the environmental conditions (MacArthur and Pianka 1966) and/or expression of individual personality consistent with time and across conditions (Sih et al. 2003, 2004; Brodin 2008; Conrad et al. 2011).

The relationship between the activity of animals and environmental conditions shows that level of activity correlates with geographical (Liebgold and Dibble 2011), seasonal (Valdimarsson et al. 1997; Benito et al. 2015) and diurnal variability (Alanärä et al. 2001; Brännäs 2008). The activity of animals is further influenced through mutual interactions, and the level of activity reflects the social status of an individual (Gómez-Laplaza and Morgan 2003). Large and dominant individuals are active at times suitable for food intake and when predation risk is low, whilst subdominant individuals are diurnally active (Metcalfe et al. 1999; David et al. 2007) and have larger home ranges (Nakano 1995). Groups containing familiar individuals are more active and better at resource exploitation and predation avoidance compared with unfamiliar conspecifics (Höjesjö et al. 1998; Griffiths et al. 2004; Liebgold and Dibble 2011). Territory holders compete for the territory more actively with intruders than neighbours (Booksmythe et al. 2010), and active individuals more effectively defend the territory than less active conspecifics (Sikkel and Kramer 2005).

Consistent individual differences in activity have been observed for individuals in ecological situations in which the behavioural syndromes are measured (Réale et al. 2007; Conrad et al. 2011; Sih et al. 2012), e.g. in brook charr Salvelinus fontinalis (M.), suggesting two alternative feeding strategies based on high activity and consistent behaviour and/or low activity and plastic behaviour (McLaughlin et al. 1992; Wilson and McLaughlin 2007; Farwel and McLaughlin 2009). Activity is often a substitutive demonstration of an individual status, e.g. stress. Øverli et al. (2005) described the consistent behavioural traits of individuals in a group of rainbow trout Oncorhynchus mikkys (W.) selected for low and high cortisol response to stress, showing faster acclimatisation to novel environments and decreasing activity in the presence of intruders with low cortisol responses. Animals show decreased activity in the stressful presence of familiar predators (Brodin and Johansson 2004; Brodin et al. 2006) and/or dominant conspecifics (Rudolf 2007). Furthermore, different levels of activity accompanied with stress can be expected for residents and intruders, reflecting differences in motivation that can be won and/or lost (Wolf et al. 2007); intruders are expected to be more active, as these individuals are bolder, more exploratory and/or aggressive (Cote et al. 2010, 2015).

The aim of the present study was to analyse the movement activity in European catfish Silurus glanis (L.) as a response to multiple stressful factors. Europe’s largest freshwater predator (Copp et al. 2009) lives in groups (Boulêtreau et al. 2011), shows variability in movement activity across seasons and diurnal rhythms (Slavík et al. 2007), and displays individual differences in energy consumption (Slavík and Horký 2012) which are higher when fish are exposed to stressful contact with unfamiliar individuals (Slavík et al. 2011).

We examined the effects of individual characteristics (body mass), and social (familiar vs. unfamiliar, resident vs. intruder) and environmental factors (resource availability) on the movement activity of juvenile European catfish. We assumed that (1) level of activity decreases with increasing body mass, because large individuals are typically dominant and thereby exposed to lower social stresses, (2) familiar individuals will be more active because familiarity often reduces stress and prolongs resource exploitation, (3) the activity of residents will be reduced in the presence of intruders according to stress-induced intrusion, and (4) shelters are highly valuable resources; hence, their availability results in decreased activity.

Materials and methods

Study organisms used

The experiment was prepared and conducted according to valid legislative regulations (Law no. 246/1992, § 19, art. 1, letter c, filed with permit no. 26758/ENV/10-1092/620/10-PP6, registered by the Ministry of Environment of the Czech Republic). The fish used in the experiment were hatchery-reared juvenile catfish. Two shoals of European catfish unfamiliar with each other were obtained from different local fish suppliers (the first shoal from the Czech Fishery, Rybářství Třeboň, and the second shoal from the Rybářství Nové Hrady, Czech Republic) to ensure that the individuals belonging to the distinct shoals had never been in contact. A total of 1600 equal-sized fish (800 from each shoal) were transported from the hatcheries to the laboratory at the age of 4 months. The fish were transported under stable conditions in oxygenated tanks in an air-conditioned loading space, and the transport lasted approximately 2 h. No effect of the transport was found on the health or mortality of the fish.

The fish were subsequently kept in 4 separate holding tanks (1000 L each, initial density 3.8 kg m−3, i.e., 400 individuals per tank) for 6 weeks prior to the start of the experiment. Each holding tank (A, B, C, and D; Fig. 1) contained individual shelters made of pipes (diameter 5 cm; length 20 cm) that were interconnected to form a honeycomb-like structure (for details, see Slavík et al. 2012). All individuals could inhabit their own shelter, as the number of shelters was higher than the number of fish. Hence, the frequency and intensity of aggressive behavior were low, and the experimenter did not need to intervene to prevent the escalation of interactions to injurious fighting. The fish were fed ad libitum on food pellets (Biomar Group, Denmark, http://www.biomar.com) distributed across the whole tank, providing free access to food to all individuals twice a day. The fish were kept under a natural photoperiod, maintaining the same regime to which they had become accustomed in the hatchery. The water was purified using biological filters with an integrated UV steriliser (Pressure-Flo 5000; Rolf C. Hagen, http://www.lagunaponds.com). The water temperature, dissolved oxygen and pH were controlled automatically (HOBO data logger; Onset Computer, Bourne, MA, USA).

Fig. 1
figure 1

Experimental design and the artificial stream. Repetitions of trials involved the regular rotation of European catfish Silurus glanis (L.) from different holding tanks used as residents or intruders to minimise the ‘fish origin’ effect on the results

Full size image

The fish were tagged 10 days prior to the start of the experiment. The fish were anaesthetised with 2-phenoxyethanol (0.2 ml L−1; Merck KGaA, Germany) and then measured (standard length LS; mean 102 mm, range 84–128 mm) and weighed (mean body mass 9.6 g, range 4–18 g). No size differences between the shoals of fish were detected (standard length p > 0.87, n = 1600; body mass p > 0.71, n = 1600). Passive integrated transponders (PIT; Trovan ID 100, 0.1 g in air, 12 mm 2.1 mm; EID Aalten, Aalten, Netherlands) were inserted into the abdominal cavity using a syringe. This method has been successfully used in behavioural experiments (Alanärä et al. 2001). No adverse effects of PIT implantation or anesthesia were observed.

All experimental fish (1600 individuals) survived, and, after the experiment, fish were released under the control of Fish Management Authorities into the fish ponds with extensive production management.

Experimental design

The laboratory experiment was conducted between September 1 and December 6, 2011, in an oval artificial stream (see Slavík et al. 2011 for details). For the purpose of this experiment, only one-half of the stream was used (5.25 m long, 0.49 m wide and 0.32 m deep). This segment was divided into 7 subunits using 8 equidistant PIT antennae. A set of 9 interconnected pipe shelters was placed within 3 of the 7 subunits in an alternating pattern (a no-shelter unit followed by a shelter unit and so on; Fig. 1). Pipe shelters have been shown to be an important resource for juvenile catfish (Slavík et al. 2012). We simulated natural stream conditions in terms of opportunities to swim freely around the shelters, even if they were occupied. We were unable to distinguish between a fish that was directly within a shelter and a fish that was simply near the shelter. Mesh was placed over the outer antennae to prevent fish from escaping from the observed stream segment. The antennae (inner area 0.49 m × 0.25 m) were designed to serve as frames to detect fish swimming through them. The antennae were connected to a recorder that stored the detection information (PIT tag code, date, time and antenna number) in its internal memory. The handling conditions were comparable to those in the holding tanks, and the water quality and flow were controlled by 2 Pressure-Flo 5000 units (60 L/min each). This arrangement generated a visible current (0.01 m s−1) circulating through the stream; however, the fish did not have to swim continuously to maintain their positions. The average temperature, dissolved oxygen content, pH, and conductivity throughout the entire experiment were 19.51 ± SD 0.22 °C, 6.53 ± SD 0.18 mg L−1, 7.24 ± SD 0.14, and 307 ± SD 5.4 μS cm−1, respectively.

To classify ‘prior residency’, 2 phases that were identical for all treatment groups were defined. In the first phase, individuals considered to be ‘residents’ were released into the artificial stream and left there for the following 24 h (beginning at 0800 hours). The data from the first phase of the experiment were not used for further analyses. In the second phase, other individuals considered to be ‘intruders’ were introduced into the artificial stream. The second phase was the experimental phase and lasted for the following 24 h. Different ‘environmental factors’ were defined by the number of available shelters in the artificial stream (resource availability). In the unlimited shelter availability treatment group, there were 27 pipe shelters and 18 individuals (9 first phase + 9 second phase) in the artificial stream, whereas in the limited shelter availability treatment group, there were 27 pipe shelters and 36 individuals (27 first phase + 9 second phase). In each trial, the repetitions were performed by regularly rotating fish from different holding tanks as residents or intruders to minimise the ‘fish origin’ effect on the results. Individual fish were not used in the experiment repeatedly. In the familiar treatment group, the individuals added during the second phase of the experiment as intruders were familiar with the residents (i.e., they were all from the same shoal and holding tank); in the unfamiliar treatment group, they were not. Combining the different features led to 4 experimental treatment groups (Fig. 1), each of which was repeated ten times. Forty individual trial repetitions under the different ecological and familiarity conditions were conducted, resulting in the use of 1080 catfish during the entire experiment. In total, we logged over 3 million recordings of catfish movements. Subsequently, the data were reduced to a dataset containing 1080 data points (one activity value per each individual fish per 24 h, see below). These data were used in the final statistical analyses.

Statistical analyses

Statistical analyses were performed using R software v.2.11.1 (R Development Core Team, Vienna, 2010). First, the PIT transmitters were used to obtain a raw data sample consisting of more than 3 million separate measurements, each of which included the ID number of the antenna, the ID number of the fish passing through that antenna, and the time. Additional explanatory variables were used to document necessary information on each trial setting (trial repetition, resource availability, familiarity, and prior residency), a fish-specific characteristic (body mass), and certain characteristics based on the individual and the opposite group (for example, the ratio between the individual’s body mass and the mean body mass of individuals from the opposite group). These data introduced unbalanced and irregular longitudinal time profiles for every individual catfish. These profiles were used to compute a total number of antennae passes per each minute based on a one-minute grid. These values (i.e., 1440 values within 24 h) were averaged for each individual to obtain the ‘individual movement activity’ value, one per subject. The individual movement activity was used as a dependent variable. The grid averaging step excludes the dependence occurring within repeated measurements. Additionally, the 1-min grid was chosen as optimal to stabilise the variance and to normalise the data.

Several different modelling approaches were tested. The GLM regression modelling framework was then applied, and a model for Gaussian data with a logarithmic link was proposed. We used the logarithmic link rather than transforming the response using logarithms to preserve the original interpretation of the activity. The final model was selected from other competing models (all two way interactions were tested) always according to a better Akaike’s Information Criterion (AIC) value considering the model with lower AIC to fit the data better (Burnham and Anderson 1998). The AIC of the final model was 27.5 and the range of the ΔAIC comparing it’s fit with other models was 7.3–58.9.

The final model had the following form:

$$\log [{\text{EIMA]}} = \mu + \Delta_{S} + \phi_{F} + \alpha_{N} + \omega_{1} \times \log [{\text{BodyMass}}] + \omega_{2} \times {\text{grMeanBodyMass}} + [\Delta \times \alpha ]_{S \wedge N}$$

where EIMA is an Expected Individual Movement Activity value predicted by the model. The parameter μ is an intercept term for the reference category (unlimited shelters, resident fish and unfamiliarity). Additional intercept corrections for non-reference categories are introduced via, Δ S , ϕ F and α N . The model also introduces 2 slope parameters, ω 1 and ω 2, to express a dependence on the logarithm of the individual’s body mass ‘log[BodyMass]’ and the ratio between the individual’s body mass and the mean body mass of the individuals in the opposite group ‘grMeanBodyMass’. The interaction term \([\Delta \times \alpha ]_{S \wedge N}\) between shelter availability and prior residency corrects for the intercept term if the model refers to an intruder under the limited-shelter treatment group. The parameter estimates, together with the standard error estimates and the appropriate significance tests (classical t test p values common for the GLM framework), are given in Table 1. All other explanatory variables considered by the experiment but not listed in the table (for example, a factor covariate for trial repetitions) that were tested in the model building stage as non-significant or did not improve model fit according to AIC were removed from the table.

Table 1 Parameter estimates with corresponding standard errors and p values for the model
Full size table

Results

The expected individual catfish movement activity was dependent on familiarity, prior residency, and body mass as well as on the resource availability (Table 1; Fig. 2). Unfamiliar fish and intruders displayed a higher level of movement activity than familiar conspecifics and residents, i.e. familiarity and prior residency decreased the movement activity of the juvenile catfish. Movement activity generally increased under the resource availability of unlimited shelters; however, the movement activity of residents and intruders differed. The model predicted a slightly lower level of movement activity in residents than in intruders. Moreover, if the shelters were limited, this difference increased significantly to a value 1.5 times greater than that associated with unlimited shelters (Table 1; Fig. 2), i.e. n intruders further increased their activity compared with residents if the shelters were limited. Nevertheless, the movement activity of the fish was also significantly influenced by body mass and the ratio between an individual’s body mass and the mean body mass of the fish from the opposite group (see also Fig. 2). If a catfish was substantially larger than the mean body mass of the opposite group, its movement activity decreased to the minimum irrespective of the ecological conditions, familiarity, and prior residency. Conversely, we observed that the smaller an individual was relative to the mean body mass of the opposite group, the higher the influence of the treatment group was on its activity.

Fig. 2
figure 2

Expected individual movement activity (EIMA) values plotted against the logarithm of the ratio between individual body mass and the mean opposite group body mass (a) and individual body mass (b) for intruders and, analogously, for residents (c, d). Four different treatment groups are distinguished (unlimited shelters, unfamiliarity black solid line; unlimited shelters, familiarity grey solid line; limited shelters, unfamiliarity black dashed line; limited shelters, familiarity grey dashed line). A vertical line is plotted for a ratio equal to zero. This value indicates that the individual’s body mass equals the mean body mass of the opposite group

Full size image

Discussion

In the present study, laboratory observations were used to demonstrate differences in the movement activity of juvenile catfish facing stressful social and environmental conditions. The results showed that when a catfish is substantially larger than the mean body mass of the opposite group, the movement activity of the catfish is decreased to the minimum, irrespective of the social and environmental conditions. We speculated that the social and environmental factors were weak, facilitating the manifestation of individual differences with respect to the behavioural syndromes theory (Sih et al. 2003; Carter et al. 2013). Considering that a low level of movement activity represents success in competition for shelters, lower movement activity with higher body mass supports this assumption. Activity was associated with boldness (Brodin 2008; Conrad et al. 2011), and considering the fact that bold individuals grow faster (Koolhaas et al. 1999), are in better condition (Brown et al. 2007), and reach higher social status (Dahlbom et al. 2011), large catfish can be suggested as bold individuals in the present study. In contrast, smaller conspecifics with a higher level of activity might represent reactive individuals that are more sensitive to stress (Øverli et al. 2005). These results indicate that high body mass, presumably interconnected with boldness and dominance, suppresses reactions to external factors and affect the results of competition. This effect is strong, particularly in combination with the motivation to defend a valuable territory (Stokkebo and Hardy 2000; Kokko et al. 2006), represented by shelter availability in this case.

Despite information that familiarity typically increases movement activity (Liebgold and Dibble 2011), the results showed that familiarity predicted low movement activity. This finding corresponds with previous studies showing that familiar fish more effectively exploit resources (Höjesjö et al. 1998; Griffiths et al. 2004), because the recognition of social partners facilitates the avoidance of ineffective time and energy use, injury, and/or predation risk (Giraldeau and Caraco 2000; Griffiths and Ward 2011). Correspondingly, the familiarity of juvenile catfish has been associated with low movement activity resulting from the advantageous characteristic of the recognition of social partners in a group. Furthermore, the results showed that the asymmetric relationship between resident and intruder was associated with differences in movement activity, indicating a lower level of movement activity of residents with no respect to changes in resource availability during the experiment, expressed as shelter limitation. Thus, resource limitation plays a fundamental role in processes in natural systems (Murdoch 1994; Turchin 1999), and its direct influence is associated with a critical degree of resource saturation (Osenberg et al. 2002). Shelters also represent highly contested resources (Sultana et al. 2013). Ford and Swearer (2013) showed that shelter limitation was the best predictor of mortality in coral reef fish. In the present study, we examined the effect of shelter limitation as a saturated critical resource of the activity of catfish. Unlimited shelter availability induced higher movement activity in juvenile catfish. Note that experimental animals typically display an initially intensive response to a novel environment, with subsequent habituation and a gradual decline in activity (Gómez-Laplaza and Morgan 2003). In contrast, the tendency of the intruders to further significantly increase activity compared with residents when the per capita availability of shelters is low suggests that intruders also show increased activity resulting from unfavourable environmental conditions. As reported, residents are predicted to prevail in defence of highly valuable territory (Stokkebo and Hardy 2000; Humphries et al. 2006; Gherardi 2006; Takeuschi 2006). Because per capita shelter availability was the only variable environmental condition in the present experiment, we speculated that residents succeed in defence of valuable territories, i.e. the residents are more invested in shelter defence, whereas intruders are more invested in searching for shelter. This investment by the intruders has been associated with higher movement activity. The lower movement activity of residents can also be considered in terms of differences in status in the social hierarchy. For example, Rosell et al. (2008) reported the over-domination of resident root vole Microtus oeconomus males in pairwise interactions. In the present study, we observed asymmetry in the movement activity of residents and intruders: the residents showed a lower level of activity, which might represent successful territorial defence. Although the outcome of competition for resources is affected by the complexities of the various factors with less apparent structure (Kokko et al. 2006), the results indicated that competition is accompanied by a lower level of movement activity by residents than by intruders.

Thus, the results presented here are consistent with previous studies describing the relationship between social experience and animal behaviour. Correspondingly, as familiarity is beneficial for resource exploitation, we recorded a lower movement activity of familiar catfish needed for shelter occupancy. Furthermore, the asymmetric relationship between residents and intruders was accompanied by differences in movement activity, with a lower level of movement activity shown by the residents. These findings are consistent with theories predicting the success of residents defending their territories when using the advantage of prior residency and higher motivation for the defence of valuable resources. These social (familiarity/unfamiliarity, resident/intruder) and environmental factors (resources availability) have been suggested as weak, demonstrating individual differences represented as body mass. High individual body mass predicted low movement activity, irrespective of the social and environmental conditions. Thus, the results of the present study highlight a role for individual characteristics in shaping the behavioural responses of catfish to external factors.